View all changelogs

Trigger.dev v4.4.2

Input streams for bidirectional task communication, batch queue performance improvements, and security hardening.

@ericallam@matt-aitken

2 contributors

1 new feature, 2 improvements, 2 bug fixes, and 7 server changes.

Highlights

Input streams for bidirectional communication with running tasks

Tasks can now receive data while they're running via typed input streams. Define a stream schema once and use it anywhere: wait in your task, send from your backend or frontend.


// streams.ts — define once, share everywhere
import { streams } from "@trigger.dev/sdk";
export const approval = streams.input<{ approved: boolean; reviewer: string }>({
id: "approval",
});


// task — suspend until a reviewer responds (frees compute while waiting)
const result = await approval.wait({ timeout: "7d" });
if (result.ok && result.output.approved) {
await publish(draft);
}


// backend — send data to the running task
await approval.send(runId, { approved: true, reviewer: "[email protected]" });

Use .once() for a non-suspending wait, .on() to register a persistent listener (useful for cancel signals on AI tasks), or send from the frontend with the new useInputStreamSend React hook. (#3146)

Improvements

  • Increase Batch trigger processing concurrency limits: Free plan 1 → 5, Hobby plan stays at 10, Pro plan 10 → 50. (#3079)
  • Add PAYLOAD_TOO_LARGE error code so oversized batch trigger items fail gracefully with a pre-failed run instead of aborting the entire batch. (#3137)

Bug fixes

  • Fix slow batch queue processing caused by spurious cooloff on concurrency blocks and a race condition where retry attempt counts were not atomically updated during message re-queue. (#3079)
  • Fix batchTriggerAndWait variants returning unknown for run.taskIdentifier instead of the correct value. (#3080)

Server changes

These changes are included in the v4.4.2 Docker image and are already live on Trigger.dev Cloud:

  • Two-level tenant dispatch for batch queue processing, replacing the single master queue with a two-level index for O(1) tenant selection and fair scheduling regardless of queue count. (#3133)
  • Server-side input streams support: API routes for sending data to running tasks, SSE reading, waitpoint creation, and a Redis cache for fast .send()-to-.wait() bridging. Includes dashboard span support and s2-lite support for self-hosted deployments. (#3146)
  • Increase batch queue processing concurrency limits: Free plan 1 → 5, Hobby plan stays at 10, Pro plan 10 → 50. Cooloff on concurrency blocks is also disabled. (#3079)
  • Move batch queue global rate limiter from the FairQueue claim phase to the BatchQueue worker queue consumer for accurate per-item rate limiting. Add a worker queue depth cap to prevent unbounded growth. (#3166)
  • Fix a race condition in the waitpoint system where a run could be blocked by a completed waitpoint and never resumed due to a PostgreSQL MVCC issue. Most likely to occur when wait.forToken() and wait.completeToken() were called at the same moment. (#3075)
  • Gracefully handle oversized NDJSON batch items instead of aborting the stream. Oversized items are emitted as pre-failed runs with PAYLOAD_TOO_LARGE, while the rest of the batch processes normally. Also fixes invalid JSON errors on lines following an oversized chunk. (#3137)
  • Require the real user to be an admin during an impersonation session. Previously only the impersonation cookie was checked; now the admin flag is verified on every request and the session falls back to the real user if admin access has been revoked. (#3078)

How to upgrade

Update the trigger.dev/* packages to v4.4.2 using your package manager:


npx trigger.dev@latest update # npm
pnpm dlx trigger.dev@latest update # pnpm
yarn dlx trigger.dev@latest update # yarn
bunx trigger.dev@latest update # bun

Self-hosted users: update your Docker image to ghcr.io/triggerdotdev/trigger.dev:v4.4.2.

Trigger.dev v4.4.1

Ready to start building?

Build and deploy your first task in 3 minutes.

Get started now